1. Field of the Invention
The present invention relates to imaging systems, image processing methods, and image processing programs that reduce noise components in image signals captured in time series.
This application is based on Japanese Patent Application No. 2008-167864, the content of which is incorporated herein by reference.
2. Description of Related Art
In general, noise components are included in digitized image signals captured by an image pickup device and by an analog circuit and an analog-to-digital converter which are both associated therewith. A noise reduction process is performed to reduce the noise components included in the image signals to obtain high-definition images. For example, various three-dimensional noise reduction processes have been proposed in which moving images are treated as temporally-continuous image frames, and noise components that are not temporally correlated are reduced.
For example, in a three-dimensional noise reduction process, a differential signal is obtained through subtraction between the current image signal and a past image signal, and a value obtained by multiplying the differential signal between the frames by a constant factor is added to or subtracted from the current image signal, thereby reducing noise. With this three-dimensional noise reduction process, although a noise-reduction effect can be obtained, image deterioration such as residual images and tailing (hereinafter, referred to as residual image components) may occur in a moving part of the image because the differential signal includes not only noise components but also motion components.
In this case, motion information obtained from the differential signal etc. is used to calculate the above-mentioned factor. The factor is reduced for the moving part to lower the noise reduction effect to reduce residual image components.
As an example of such noise reduction processes, in Japanese Unexamined Patent Application, Publication No. HEI 6-62283, data temporally and spatially correlated to input pixel data is searched for to reduce noise. In Japanese Unexamined Patent Application, Publication No. HEI 6-62283, a nonlinear filter is used that assigns a large weight to a pixel having a value close to that of the input pixel data, assigns a small weight to a pixel having a value far from that of the input pixel data, and averages the pixel values. With this filter, a noise reduction process that corresponds to the movement of an object is performed.
As another example, in Japanese Unexamined Patent Application, Publication No. 2003-219208, the maximum value and the average value are detected among differential data from a preceding frame, the differential data being obtained for a predetermined period of time; setting information used to generate correction data and a control signal used to control the noise reduction process on and off are generated from the detected maximum value and average value; the amount of noise is detected with precision; and noise reduction is performed. In Japanese Unexamined Patent Application, Publication No. 2002-33942, the difference in pixel signal between a plurality of frames is detected as the degree of movement; the gain of cyclic noise is set based on the result; and noise reduction is performed.
As still another example, in Japanese Unexamined Patent Application, Publication No. 2005-347821, an image signal level and a movement detection result obtained from the frame difference are used to control a cyclic factor of noise, thereby performing a noise reduction process corresponding to the brightness of an image.
As still another example, Japanese Unexamined Patent Application, Publication No. 9-81754 discloses an apparatus in which the degree of movement is compared from the difference between frames; comparison results are checked by a majority decision circuit; a variation of the comparison results is corrected; the result is used as a movement signal to decide a cyclic factor of noise; and movement is detected successfully to perform noise reduction.
As described above, in the conventional three-dimensional noise reduction processes, the difference value from the current image signal is used to control the cyclic factor of noise to reduce residual image components detected when the image includes motion components.
However, the differential signal includes both the difference in motion component between images and the difference in noise component included in image signals. Therefore, when the sensitivity of detection of motion components is increased, the sensitivity of detection of the difference in noise components is decreased. The difference in noise components is thus falsely detected as motion components, and thus a good noise-reduction effect cannot be obtained. On the other hand, when the sensitivity of detection of motion components is decreased, motion components are falsely detected as noise components, leading to the disadvantage that residual image components occur in the original moving part.
Further, for example, when an image signal having a large noise-component value, such as that obtained when the gain of an input image-signal level is increased, is input, the difference in noise components is falsely detected as motion components, leading to a problem in that a noise-reduction effect cannot be obtained and a good noise reduction cannot be performed.